Seminar

Modeling drug resistance in HIV using conjunctive Bayesian networks

I will show how to understand the risk of the development of drug resistance in HIV with new statistical and mathematical tools. Statistically, we use a new class of graphical models to learn the mutational pathways along which the virus escapes from drug pressure. These models are called conjunctive Bayesian networks (CBNs); they describe the accumulation of (genetic) events with constraints on the order of occurrence. I will present a combinatorial solution to the model selection problem for CBNs and discuss how to deal with noisy data. Next, we estimate the risk of escape along these pathways. To estimate the probability that the population develops a drug-resistant mutant before extinction, we use tools from algebraic combinatorics to analyze probabilistic models on fitness landscapes. These methods are applied to two datasets where the events are HIV mutations associated with drug resistance to the protease inhibitors ritonavir and indinavir. This is joint work with Niko Beerenwinkel and Bernd Sturmfels.